Architecture Overview

System Overview

ReCloud is an advanced three-tier architecture that transforms standard coding assistants into autonomous AI orchestration engines. The system implements sophisticated multi-step workflows through intelligent tool chaining, enabling complex task execution that goes far beyond simple prompt-response interactions.

graph TB
    subgraph "User Interface Layer"
        UI[Desktop Application<br/>System Orchestrator]
    end

    subgraph "Logic Layer"
        MCP[MCP Server<br/>Tool Interface]
    end

    subgraph "Service Layer"
        API[REST API Server<br/>Workflow Engine]
    end

    subgraph "External Services"
        GEMINI[Google Gemini AI]
        FS[(File System)]
        CURSOR[Cursor IDE<br/>Client]
    end

    CURSOR --> UI
    UI --> MCP
    MCP --> API
    API --> GEMINI
    MCP --> FS
    API --> FS

Core Architecture Patterns

Dynamic Configuration System

The system implements sophisticated runtime agent configuration through integrated policy sources that enable contextual agent adaptation based on user preferences and task requirements.

Protocol Integration

Complete integration with modern IDEs through standardized communication protocols for seamless tool exposure and context preservation.

Hybrid Execution Model

Intelligent execution routing across three processing targets:

Remote Execution: Cloud-based AI processing for complex tasks
Local Execution: Hardware-accelerated processing on user systems
Hybrid Execution: Optimized combination of remote and local processing

Declarative Configuration Framework

Configuration framework that enables declarative workflow specification without requiring code modifications.

System Components

1. Backend API Server

Purpose: Enterprise-grade REST API server for AI task processing and complex workflow execution.

Technology Stack:

Node.js + TypeScript + Express.js for robust API development
Google Generative AI integration for advanced AI processing
Comprehensive data validation and type safety
Template processing for dynamic content generation

Key Capabilities:

High-performance REST API with comprehensive endpoint coverage
Advanced Google Gemini AI integration with multiple model support
Workflow execution engine for complex task orchestration
Enterprise-grade file processing and data transformation capabilities

2. MCP Logic Server

Purpose: Advanced server implementing intelligent tool processing with unified interface design.

Technology Stack:

Node.js + TypeScript for type-safe development
Model Context Protocol SDK integration
Dynamic configuration engine
Intelligent execution routing

Innovative Features:

Unified Tool Interface: Single tool with comprehensive subcommand system
Dynamic Configuration: Runtime directive creation from multiple policy sources
Intelligent Routing: Advanced hybrid execution with optimal resource utilization
Context Preservation: Complete state management across complex workflows

3. Desktop Orchestrator

Purpose: Cross-platform desktop application providing complete system orchestration and user experience management.

Technology Stack:

Electron framework for native cross-platform desktop applications
Embedded server distribution for seamless operation
Persistent configuration management

Core Functions:

Complete System Orchestration: Full server lifecycle and resource management
Advanced Configuration: Sophisticated agent behavior profile management
Seamless IDE Integration: Automatic IDE configuration
Real-time Monitoring: Comprehensive health checks and status reporting
Intelligent Tool Management: Project-level tool activation with persistence

Data Flow Architecture

Primary Execution Flow

sequenceDiagram
    participant Cursor as Cursor IDE
    participant TrayApp
    participant MCP
    participant API
    participant Gemini

    Cursor->>TrayApp: Tool request
    TrayApp->>MCP: Execute command
    MCP->>MCP: Apply configuration
    MCP->>API: Remote processing
    API->>Gemini: AI processing
    Gemini-->>API: AI response
    API-->>MCP: Formatted result
    MCP-->>Cursor: Protocol response

Configuration Assembly Process

sequenceDiagram
    participant TrayApp
    participant Environment
    participant ConfigBuilder
    participant MCP

    TrayApp->>Environment: Set configuration
    Environment->>ConfigBuilder: Read settings
    ConfigBuilder->>ConfigBuilder: Build configuration
    ConfigBuilder->>MCP: Apply configuration
    MCP->>MCP: Use configuration

Workflow Engine Architecture

Declarative Configuration System

The workflow engine uses a configuration framework that enables declarative workflow specification:

name: analyzeSentiment
execution_target: remote
model: "gemini-2.5-flash"
parameters:
  - name: "text"
    type: "string"
    required: true
step_types:
  - name: "ai_prompt"
    handler: "ai_prompt"
    prompt_template: "Analyze sentiment: {{flow_input.text}}"
flow:
  - step: "analyze_sentiment"
    type: "ai_prompt"

Step Handler System

The workflow engine implements a modular handler system:

AI-powered text processing with templating
External API integrations with authentication
File system operations with security validation
Cloud execution environments
Internal service orchestration

Execution Routing

Intelligent routing between execution environments:

flowchart TD
    A[Receive command] --> B[Check configuration]
    B --> C{Execution target?}

    C -->|remote| D[Remote API call]
    C -->|local| E[Local handler]
    C -->|hybrid| E

    D --> F[Return result]
    E --> G[Execute locally]
    G --> F

System Dependencies

Component Relationships

Desktop Application
    └── Embedded MCP Server
        └── Tool routing system

MCP Server
    └── Backend API (HTTP orchestration)
        └── Google Gemini AI (processing)

Backend API
    └── Google Gemini AI (processing)
    └── File System (persistence)

External Dependencies

Google Gemini AI: Advanced AI processing and reasoning
Cursor IDE: Primary IDE integration
File System: Results storage and caching infrastructure

Performance & Scalability

Optimization Features

Multi-level Caching: Intelligent cache hierarchies for configuration and execution data
Resource Management: Efficient memory and processing resource utilization
Concurrent Processing: Parallel execution support for multi-step workflows
Load Distribution: Intelligent routing across available processing resources

Monitoring & Observability

Execution Metrics: Comprehensive tracking of workflow performance and latency
Health Checks: System status monitoring and automated recovery
Logging Integration: Structured logging with configurable verbosity levels
Performance Analytics: Detailed insights into system utilization and bottlenecks

ReCloud provides the robust architectural foundation for advanced AI orchestration, enabling complex multi-step workflows with enterprise-grade performance and reliability.